What Is the 10-20-70 Rule for AI?

I learned this the slow way, inside real businesses, not on a slide.

Where the rule comes from

The 10-20-70 breakdown was popularized by BCG’s research on AI transformations. Once you have done a few of these, it stops feeling like a statistic and starts feeling like a law of nature. The math is not precise to the decimal. The point is the proportion. The intelligence is cheap and getting cheaper. The plumbing is solvable. The people are everything.

Why the 70 percent is the whole game

When an AI project fails, the post-mortem almost never says “the model was not smart enough.” In three years of putting AI to work inside real organizations, I have not seen that happen once. What I see instead:

• Nobody redesigned the workflow, so the AI got bolted onto a broken process and made it faster and more broken.
• The team did not trust it, so they quietly kept doing the old thing on the side.
• No one owned it after launch, so it drifted, went stale, and got abandoned.
• Leadership wanted the output without changing how anyone actually works.

That is all 70 percent work. It is change management. It is training. It is trust. It is redesigning the way a day flows so the tool has somewhere to live. None of it shows up in a demo, which is exactly why it keeps getting underfunded.

Why this is good news

Here is the part I love. If 70 percent of AI success is human, then your advantage is human too. You do not need the biggest model or the deepest pockets. You need to understand your own work well enough to put AI in the right place. That is a level playing field, and it favors the people who actually know how the work gets done.

This is the core of how I think about implementation. The gap most organizations have is not a lack of technology. It is sequence. They reach for the 10 percent first because it is exciting, and they save the 70 percent for last, if they get to it at all. Flip that order and everything changes. That is the whole idea behind the Sequence Model.

What a small business should actually do with this

If you run a small business and you are staring down AI, do not start by shopping for tools. Start with the 70 percent.

1. Map how the work really happens now. Not the org chart. The actual flow of a day.
2. Find the noise. The repetitive, draining, low-judgment tasks that eat your people’s hours.
3. Put AI there first, where it buys back time without asking anyone to trust it with the hard calls.
4. Train and document, so it survives past the first excited week.
5. Then, and only then, worry about which model.

That is foundation before growth, which is the spine of the Mission-Driven AI Stack. Build the base, then climb.

The deeper point

The 10-20-70 rule is usually taught as a project-management warning. I think it is something bigger. It is proof, sitting right there in a consulting firm’s data, of the thing I keep saying. AI does not make the human less important. It makes the human the entire point. The technology is the cheap part now. Judgment, context, trust, and care are the expensive part, and they always will be.

So when someone tells you AI is going to replace your people, you can hand them the math. Ninety percent of the value was never in the machine.

If you want help finding your own 70 percent, that is exactly what an AI Strategy Session is for. We map your work, find the right place to start, and build a plan that puts the human back at the center.

AI does not make the human less important. It makes the human the entire point.

FAQ